Skip to content

Cognitive Disequilibrium

Principle · All principles
Evidence · 30 claims (15 for, 12 mixed, 3 against) · 31 studies (13 causal, 5 quant-synthesis, 5 review, 3 theoretical, 2 qualitative, 2 design, 1 associational), q1–q4 · 8 of 31 report an effect size · 25 claims rest on one study

Conditional relationship

For a learner who already holds an explanation of a phenomenon and states it with some confidence, but whose explanation is inaccurate or too simple for the target, meeting evidence that the explanation cannot account for, and then being helped to resolve the mismatch, is expected to improve conceptual-understanding of the target principle more than presenting the correct account alone. The knowledge type is mainly principle (causal models and explanations, most often in science), sometimes concept. The outcome is the revised explanation on a probe that the old explanation would answer wrongly; whether the revision holds at delayed-retention and in unfamiliar contexts (near-transfer, far-transfer) is a separate question, and the evidence here says less about it.

The relationship is conditional on four things a page cannot settle in advance: that the learner's starting response actually shows the inaccurate model (a guess, a gap or a misread prompt is not a misconception); that the contradiction bears on the target concept rather than on an incidental feature; that the learner notices and experiences it as a conflict, which a staged contradiction does not guarantee; and that support for resolving it follows (explanation prompts, discussion, an explicit statement of the better model). Contradiction without those last two may leave the model unchanged, or produce confusion and defensiveness instead.

Two converted pages hold narrower or neighbouring models. Activation covers making prior knowledge visible before instruction, including the case where the activated idea is a misconception and the benefit depends on the instruction confronting it; this page owns what happens after that idea is visible: the confrontation and its resolution. Problem-Based Learning holds productive failure, a problem attempted before instruction; that is one way of producing a mismatch, scoped to conceptual knowledge and transfer with instruction following, and its model is not repeated here. No pattern page instantiates this principle yet.

Default design, while the relationship is untested

The page's earlier guidance, kept as a concrete default a designer can act on and revise. Each step is a design proposal; the evidence for it is stated beside it, and no claim here tests the sequence as a whole.

  1. Elicit and record the learner's model first. Ask for a committed prediction or explanation of a case the target concept decides, and record the learner's confidence. A conflict needs something to attach to. Evidence: the conceptual-change syntheses below test instruction built this way, not this step on its own; that confidently held wrong answers are corrected better rests on a claim about answers corrected by feedback, not about explanatory models.
  2. Choose a contradiction tied to the target concept: a discrepant event, a data set, a case the learner's explanation predicts wrongly. Untested here; from the earlier page.
  3. Show the disconfirming evidence soon after the commitment, while the prediction is still active. Untested here; the prediction element and the claim pages' Discussions propose it.
  4. Check that the conflict was noticed. Ask what was expected and what happened, before explaining. In one experiment, staged contradictions raised learning only where learners reported being confused by them (see below).
  5. Support the resolution. Prompt the learner to explain the mismatch, discuss rival explanations, and then state the better model explicitly, naming the old idea and why it fails. Evidence: refutation text (naming the misconception, then refuting it) has review and experimental support for comprehension and recall of the corrected facts; self-explanation is under Further evidence, not read against this model.
  6. Keep revision safe. Make being wrong a normal, non-punitive event, so the learner revises rather than defends. Untested here; from the earlier page.
  7. Judge on the old model's home ground, at the horizon the objective sets: a delayed, differently worded probe where the old explanation would give a different answer. A correct answer right after the lesson is not yet evidence that the old model is gone. Untested here; the misconceptions claim's Discussion raises reversion, and no recorded study measures it.

How far each neighbouring claim may be carried is said under each one below.

Fitting the design to a situation

Ask for four facts when the brief omits them: is there a known, recorded wrong model for this topic (or a way to elicit one)?, will there be support to resolve the conflict once it is noticed?, do learners work alone, in pairs or from text?, and when will the revised model be tested?

If the brief says… Then change… Basis
The topic is science with well-known misconceptions (force and motion, heat, natural selection) Build the unit as predict, confront, resolve: elicit a committed prediction, show the case it gets wrong, then teach the better model explicitly Conflict-based instruction improves science conceptual learning [+M]
The material is mainly read (textbook, online text) Use refutation text: state the common misconception, say that it is wrong, and give the scientific account with its reason Instruction that confronts misconceptions produces conceptual change more often than exposition [+M]; Refutational text order enhanced recall in one experiment [+W]
Learners answer questions with a stated confidence (quizzes, clickers, polls) Ask for confidence with each answer, and give immediate, clear, non-punitive correction; confident errors are then corrected best High-confidence errors lead to better retention after correction [+M]
Learners will resolve the conflict in pairs or groups (for example at a simulation) Add an individual step: each learner writes what they expected, what happened and how they now explain it; peer conflict changed conceptions only for students who reflected on and rebuilt their own Peer conflicts worked only for students prepared to reflect [~M]; Co-construction needed personal sense-making for stable change [~W]
The group looks engaged and contributions look balanced Still check each learner's model on an individual probe; equal, mutual engagement did not ensure conceptual change High equality and mutuality were not sufficient for conceptual change [~W]
Novices meet the problem before any instruction, with no support Add scaffolding to the attempt so that confusion and surprise lead somewhere; unscaffolded attempts produced more confusion, and emotions related to transfer differently by condition Emotion dynamics predict learning in a way that depends on scaffolding [~M]
Learners are getting most practice items wrong and say they are confused Lower the difficulty or give support now rather than more contradiction; confusion with repeated errors predicted worse later performance Confusion amplifies the negative effect of incorrect trials [-W] (exploratory)
The target is a developmental reasoning skill and the brief asks how big a challenge to set Pitch the discrepant task one level above the learner's current level; two levels above produced no more advance, and all advances were one level One-step advance regardless of discrepancy size [~W] (one 1975 spatial-reasoning study)
The revised model must hold for months Plan a delayed probe the old model would answer wrongly, and revisit; in one physics study 7 of 14 students showed no change at post- or delayed test Force-and-motion conceptions were very difficult to change [~W]
Learners will discuss in cooperative groups (for example college chemistry) Prepare group leaders or assign roles so the discussion stays on the conflicting ideas; poor leadership undermined it Conceptual change strategies in cooperative groups reduced misconceptions [~W]
Learners are anxious about being wrong, or the setting is public Collect predictions privately or anonymously and say that wrong predictions are expected untested proposal

Observation, state and explanation

What a designer can observe is a response under stated conditions: a prediction and its stated confidence, an explanation in the learner's words, what the learner says or does when the evidence contradicts it, and the explanation given afterwards. "Holds a misconception", "is in disequilibrium" and "has revised the model" are inferred states. Surprise, confusion, silence and argument are observable, but none of them shows which state the learner is in. Record the case used, the prompt, the confidence asked for and every hint or explanation given. The rows below are proposals, untested with learners.

Response Plausible explanations Observation that could change the interpretation
Confident wrong prediction; after the discrepant result, says "that was a trick" or keeps the prediction A coherent model the learner protects; the result is read as an exception or a fault in the demonstration; low trust in the source; face-saving in front of peers Ask privately for the learner's account of the result, and offer a second case the same model also predicts wrongly; a learner who explains both away holds the model, one who hesitates on the second is beginning to doubt it.
Shows surprise, then gives the taught explanation Revised model; the taught words repeated without a changed model; compliance with the teacher A delayed, differently worded case where the old model and the new one predict different outcomes, answered without the lesson's vocabulary as a cue.
Shows no surprise at the contradicting evidence Did not hold the predicted model (the prediction was a guess); did not notice the mismatch; does not see the case as bearing on the idea Compare the stated confidence with the response; ask the learner to say what they expected and what happened; repeat with a case where the mismatch is harder to miss.
Becomes confused and stops, or gives up Conflict without enough support to resolve it; the case carries too much else to process; low confidence in the domain Offer a structured prompt (compare your prediction and the result; which part of your explanation fails?) and see whether work resumes; note whether the difficulty is in the concept or in reading the case.
Revises the explanation without any conflict having been staged Learned from the explanation or from discussion; already partly held the better model; reflection prompted by the task itself Look at the first response again; compare learners who met the conflict with learners who did not, keeping all exposures in the record. In one qualitative study some students changed without any peer conflict.

These probes can shift confidence between explanations; none identifies a cause on its own, and each probe is a further exposure that can change the state.

Evidence and qualifications

No claim in the wiki tests the relationship as stated: none compares, for learners whose starting model was recorded, a conflict-and-resolution sequence against the same content presented without the conflict, on a probe of the old model at a delay. The claims below test conflict-based instruction as a whole, one form of it, or the conditions under which conflict does nothing.

  • Conflict-based instruction improves science conceptual learning, though staged contradictions helped only learners who reported being confused, and no study isolates disequilibrium as the mechanism [+M]: three entries. A random-effects meta-analysis of 218 science-education studies (18,051 students) found conceptual-change strategies produced a large improvement in learning (g = 1.10; g = 0.93 after a robust Bayesian adjustment), and the cognitive-conflict family did as well (g = 1.10, k = 150), with a wide prediction interval (0.19 to 2.38). An earlier meta-analysis of reading- and science-education interventions on misconceptions reported that the effective ones shared an element of producing conceptual conflict; no effect size or study count is recorded. Both were read as abstracts. The syntheses measure learning outcomes, not whether disequilibrium was felt, so they support conflict-based instruction as a package and do not isolate the conflict or say what the comparison conditions were. The third entry qualifies it: in a within-subject experiment with 32 undergraduates (adult, lab), contradictions staged between two tutoring agents raised self-reported confusion but did not on their own raise posttest scores; in a median-split follow-up, learners scored higher after contradictions only in the cases where they reported high confusion. The study gave no support for resolving the conflict. A judge failure on this claim was reviewed and dismissed; its entries are otherwise not settled by the text available.
  • Instruction that confronts science misconceptions (conceptual-conflict interventions, refutation text) produces conceptual change more often than standard exposition [+M]: the same 1993 meta-analysis (so not an independent second study) and a review of two decades of refutation-text research, which found that reading text that names a common science misconception and argues against it was more likely to produce conceptual change than traditional expository text on the same topic, with no developmental pattern. Both read as abstracts; no effect sizes. It supports step 5's explicit refutation and the premise that presenting the correct account alone often does not dislodge a misconception; it does not test a staged discrepant event, and its claim page says its own strength tags await citation-level support.
  • In one experiment, refutational text (misbelief stated first, then fact) enhanced free recall relative to traditional text and free and cued recall relative to inverted refutational text, but its reported 14-day recognition advantage was over inverted refutational text only [+W]: one experiment on students' inaccurate science beliefs. Refutational text beat traditional text on free recall of the facts, beat inverted refutational text on free and cued recall, and beat inverted refutational text on 14-day recognition. The outcome is recall and recognition of facts, not an explanatory model; the learners and setting are not described on the claim page. Its title says 14-day retention relative to traditional text as well, which its subclaim does not show. Not yet checked against its source.
  • High-confidence errors are more likely than low-confidence errors to be corrected on a later test after feedback. [+M]: a narrative review and one experiment (sample not stated) on the hypercorrection effect: answers given with high confidence and then corrected are remembered correctly more often than low-confidence errors, with surprise-driven attention as the proposed mechanism. The claim page says the effect depends on immediate, clear, non-punitive corrective feedback. It supports recording confidence and correcting clearly; it concerns corrected answers, not the restructuring of a model, and its material is not described on the claim page. Neither entry could be confirmed from the text available.
  • Peer conflicts did not always produce conceptual change; they appeared to work only for students prepared to reflect on and reconstruct their conceptions [~M]: a qualitative study of seven student pairs working at the computer on science (Grade 10 force and motion, per a related claim from the same study), with 0 to 8 peer conflicts per pair. Some students changed their conceptions without any peer conflict, and others met conflicts and did not change; the authors judged the crucial factor to be whether students were prepared to reflect on and reconstruct their conceptions. Checked by the judge against the full text. It limits the relationship: conflict was neither necessary nor sufficient here.
  • Emotion dynamics during problem-solving predict learning outcomes in a manner that depends on scaffolding design [~M]: a randomized experiment with 132 university students (adult, novice to data science) comparing failure-driven scaffolding, success-driven scaffolding and unscaffolded problem solving before instruction. Unscaffolded problem solving produced more confusion than either scaffolded condition (d = 0.24–0.43); both scaffolded conditions produced more surprise than it did. Within conditions, the same emotions (disgust, contempt) correlated with transfer scores in opposite directions depending on the condition; those correlations are observational. It qualifies any reading of confusion as a marker of productive disequilibrium: what a given emotion means for learning depended on the design around it. One study, one domain; not yet checked against its source.

Taken together: conflict-based conceptual-change instruction has large average effects in science education, with wide variation and abstract-only reading; naming and refuting a misconception has review and experimental support for comprehension and fact recall; and conflict on its own, whether staged by agents or arising between peers, did not reliably change learning unless learners experienced it and were prepared to work on it. Learners, settings and outcomes differ across these claims, so do not rank them by effect label.

How far each may be carried: the syntheses are about science-education instruction packages, not a single discrepant event, and not other domains; the refutation-text claims concern reading, not demonstrations or discussion; the hypercorrection claim concerns corrected answers, so it says nothing about explanatory models; the agent-contradiction and emotion studies are single-session or single-domain studies with university students. None establishes a delayed benefit on an explanatory probe, and none tests the default design's checking, timing or safety steps.

The evidence by question

Each question below links a general claim that gathers the studies behind it into one table, one row per distinct study, with each study's design and effect as recorded and no averaged verdict (eval/synthesis/cognitive-disequilibrium.json, scripts/synthesize_claim.py). The count is of distinct studies across the claim and the narrower claims that bear on it.

Further evidence, not yet read against this model

Claims this page cited before it was rewritten as a conditional model. They are evidence near the relationship, but none has yet been re-read against the model above; each says how far its own sources have been checked.

Objective and learner-valued goal

The designer's objective names the model to be revised, the probe that would tell the old model from the new, and the horizon. Elicit separately what the learner wants from the topic: to get the exam question right, to understand why something they have seen happens, not to look wrong in front of others. Agreement and divergence are both observations; do not infer readiness to revise from a willingness to answer.

For example, a secondary physics student who predicts that a heavier ball will land first may value getting the test item right, while the designer's objective is an explanation of motion that holds on unfamiliar cases a month later. A dropped-ball demonstration followed by the correct answer may satisfy the learner's aim and leave the designer's untested; a delayed case worded differently, where the old idea and the new one predict different results, tests the objective. If the learner's aim is mainly not to be seen getting it wrong, a public prediction may provoke defence rather than revision, and a private prediction is the better first move.

What would revise this model?

The model should weaken if controlled comparisons, for learners whose starting model was recorded, show that a conflict-and-resolution sequence does no better on a delayed explanatory probe than a clear explanation of the correct model alone, or no better than refutation alone without a staged conflict. If gains follow the resolution support whether or not a conflict was staged, as some students in the peer-conflict study suggest, the weight belongs on the support, not the contradiction. If the large synthesis averages turn out to rest on comparisons with weak control instruction, they no longer bear on the conflict itself. Do not protect the model by calling every unchanged learner "not truly in disequilibrium" or "not ready to reflect" after the fact; record what was observed before the result is known.

Surprise or confusion during the lesson, a correct answer immediately after it, a revised explanation at a delay and use of the model in a new context are separate claims. The present evidence does not establish how strong a contradiction should be, how soon resolution should follow, which learners respond with defence rather than revision, or whether a revised model persists.

Claims

  • Error Analysis — cognitive disequilibrium often begins by surfacing and examining an incorrect but plausible model
  • Cognitive Flexibility — once the original model is destabilized, learners may need to represent the concept in more than one way
  • Cognitive Activation — disequilibrium is one route to deeper cognitive engagement

Examples

Illustrative

Discrepant Events — A demonstration produces a result opposite to what learners predicted, forcing them to reconcile expectation and evidence.

Prediction-then-observation routines — Learners commit to an explanation before seeing data or a result, then must account for the mismatch when their prediction fails.

Historical or scientific misconception challenges — Students compare a common but flawed explanation against a stronger one using concrete evidence that the initial account cannot explain.

Key Sources

  • Posner, G. J., Strike, K. A., Hewson, P. W., & Gertzog, W. A. (1982). Accommodation of a scientific conception: Toward a theory of conceptual change. Science Education, 66(2), 211-227. https://doi.org/10.1002/sce.3730660207
  • Limon, M. (2001). On the cognitive conflict as an instructional strategy for conceptual change. Learning and Instruction, 11(4-5), 357-380. https://doi.org/10.1016/S0959-4752(00)00037-2
  • Bautista, R. G. (2015). Students' affordance of teleologic explanations and anthropomorphic language in eliciting concepts in physics. Journal of Technology and Science Education, 5(1), 31-40. doi:10.3926/jotse.137
Page metadata
Field Value
type principle
id cognitive-disequilibrium
aliases account-for-existing-ideas-correct-teleologic-language
title Cognitive Disequilibrium
description For a learner who holds a confident but inaccurate explanation, meeting evidence it cannot account for and then being helped to resolve the mismatch is expected to improve conceptual understanding more than the correct account alone, with weak evidence that the revision lasts or that conflict without support helps.
status review
canonical True
generated by claude/unspecified on 2026-10-02
trust unverified — no one has confirmed this page's content
sources 2 — posner-1982, bautista-2015

Frontmatter is the page's machine-readable record — see the Schema & Guide.